Intra-tumoral heterogeneity of KRAS and BRAF mutation ... - Hindawi

61

Analytical Cellular Pathology 34 (2011) 61–66 DOI 10.3233/ACP-2011-0005 IOS Press

Intra-tumoral heterogeneity of KRAS and BRAF mutation status in patients with advanced colorectal cancer (aCRC) and cost-effectiveness of multiple sample testing1 Susan D. Richmana , Philip Chambersb , Matthew T. Seymoura , Catherine Dalya , Sophie Granta , Gemma Hemmingsa and Philip Quirkea,∗ a The Sections of Pathology and Tumour Biology and Oncology, Leeds Institute of Molecular Medicine, University of Leeds, Leeds, UK b Leeds Institute of Molecular Medicine, Genomics Support, University of Leeds, Leeds, UK

Abstract. KRAS mutation status is established as a predictive biomarker of benefit from anti-EGFr therapies. Mutations are normally assessed using DNA extracted from one formalin-fixed, paraffin-embedded (FFPE) tumor block. We assessed heterogeneity of KRAS and BRAF mutation status intra-tumorally (multiple blocks from the same primary tumor). We also investigated the utility and efficiency of genotyping a ‘DNA cocktail’ prepared from multiple blocks. We studied 68 consenting patients in two randomized clinical trials. DNA was extracted, from ≥2 primary tumor FFPE blocks per patient. DNA was genotyped by pyrosequencing for KRAS codons 12, 13 and 61 and BRAF codon 600. In patients with heterogeneous mutation status, DNA cocktails were prepared and genotyped. Among 69 primary tumors in 68 patients, 7 (10.1%) showed intratumoral heterogeneity; 5 (7.2%) at KRAS codons 12, 13 and 2 (2.9%) at BRAF codon 600. In patients displaying heterogeneity, the relevant KRAS or BRAF mutation was also identified in ‘DNA cocktail’ samples when including DNA from mutant and wild-type blocks. Heterogeneity is uncommon but not insignificant. Testing DNA from a single block will wrongly assign wild-type status to 10% patients. Testing more than one block, or preferably preparation of a ‘DNA cocktail’ from two or more tumor blocks, improves mutation detection at minimal extra cost. Keywords: KRAS, BRAF, heterogeneity, colorectal cancer

1. Introduction The drug treatment of advanced colorectal cancer (aCRC) has seen recent advances; first newer cytotoxic drugs such as oxaliplatin (Ox) and irinotecan ∗ Corresponding author: Prof. Philip Quirke, Pathology and Tumour Biology, Leeds Institute of Molecular Medicine, Wellcome Trust Brenner Building, Leeds University, St James’s University Hospital, Leeds LS9 7TF, UK. Tel.: +44 (0)113 343 8407; Fax: +44 (0)113 343 8431; E-mail: [email protected]. 1 Previous Presentation: Pathological Society of Great Britain and Ireland Summer meeting 2009, Cardiff 30th June–3rd July.

(Ir) [1–3]; more recently, monoclonal antibody (mAb) therapies targeting vascular endothelial growth factor (VEGF) [4] and the epidermal growth factor receptor (EGFr) [5]. Anti-EGFr-mAb therapies, cetuximab and panitumumab, have now been approved for the treatment of aCRC. Although beneficial to some patients, these drugs are costly [6], they cause significant toxicity [7], and if used unselectively may provide minimal or even negative [8, 9] net benefits. Therefore, attention has rightly focused on identifying predictive biomarkers for patient selection. The first of these, now included in amended drug licenses, is the KRAS oncogene.

2210-7177/11/$27.50 © 2011 – IOS Press and the authors. All rights reserved

62

S.D. Richman et al. / Intra-tumoral heterogeneity of KRAS and BRAF mutation

KRAS protein is involved in one of several EGFr signal transduction cascades. It is activated following ligand binding to the extracellular domain of EGFr, triggering downstream events including activation of the mitogen-activated protein kinases (MAPK). Mutation hotspots in KRAS codons 12, 13 and 61, lead to constitutively active KRAS protein, and thus EGFrindependent activation of the MAPK pathway [10–12]. BRAF, downstream from KRAS in the MAPK pathway, is subject to an activating mutation, in codon 600. It is known that activated KRAS (KRAS-mut) confers clinical resistance to anti-EGFr-mAbs. Whilst patients with wild-type KRAS (KRAS-wt) cancers have, in some trials, benefited from cetuximab or panitumumab, those with KRAS-mut cancers have not, and in some trials have been harmed by them. Recent data suggest that activating mutations of NRAS or BRAF, also confer anti-EGFr-mAb resistance [13]. Their lower prevalence makes this harder to confirm within single trials, but in three recently-reported grouped series including over 500 patients receiving cetuximab or panitumumab-based treatments there were 35 patients with KRAS-wt, BRAF-mut tumors, none of whom responded to therapy [13]. Similarly, no responses have been seen in patients with NRAS-mut. KRAS and BRAF are also prognostic factors independent of treatment. We previously showed, in 711 aCRC patients treated without anti-EGFr therapy, that mutation in either oncogene is negatively prognostic for survival, with KRAS-mut giving a hazard ratio (HR) of 1.24 (95% CI 1.06–1.46; p = 0.008) and BRAF-mut giving HR = 1.82 (95% CI 1.36–2.43; p < 0.0001) [14]. In a recent report of 516 aCRC patients randomized to receive chemotherapy and bevacizumab with or without cetuximab, BRAF-mut was associated with inferior outcomes in both arms [15]. Currently, molecular testing usually involves DNA extraction from a single tumor tissue block, followed by a mutation-specific PCR-based assay or sequencing of the relevant codons. Sensitivity and errors of these assays in different laboratories may be significant, and are being addressed by international quality assurance programs [16]. Here we assess a second potential source of error: tumor heterogeneity. Using material from patients in randomized clinical trials, we ask how frequently KRAS or BRAF mutation status is heterogeneous within tumors, and therefore how representative is a single test as currently practiced. We also investigate the options of multiple testing, or of

performing a single analysis using a ‘DNA cocktail’ extracted from multiple blocks per tumor. Based on heterogeneity levels in this study, we have estimated the clinical and cost consequences of these different approaches.

2. Methods 2.1. Patients Material was available from aCRC patients in two large UK National Cancer Research Institute randomized clinical trials: FOCUS and PICCOLO. In each, separate consent was obtained for the use of surplus stored pathological material for research. From the trial biobank of over 2000 patients, patients were selected at random for this study provided they had at least two separate cancer-containing blocks from different areas of the primary tumor. Sample size was determined by the availability of material at the time of study. FOCUS (Fluorouracil, Oxaliplatin and CPT11 [irinotecan]: Use & Sequencing) involved 2135 patients randomized at 60 centers between 2000 and 2003, comparing different sequences of first- and second-line chemotherapy for aCRC [17]. Formalinfixed, paraffin-embedded (FFPE) tumor blocks were retrieved retrospectively. PICCOLO (Panitumumab, Irinotecan and Cyclosporin in Colorectal cancer therapy) opened in December 2006, comparing different second-line therapies. Unlike FOCUS, PICCOLO randomization differs according to KRAS status, so tumor samples are obtained prior to randomization. 2.2. Study plan First, in line with standard practice, a single primary tumor sample from each patient was analyzed for mutations in KRAS codons 12, 13 and 61 and BRAF codon 600. The rate of mutations at each codon was assessed. For each patient, all further available primary tumor blocks were sampled and analyzed in the same way as the first sample. These results were then compared with the initial result and with each other to assess the rate of intratumoral heterogeneity. As an internal control, one series of samples was duplicated, to assess intra-lab variability. In patients identified as having intratumoral heterogeneity, a ‘DNA cocktail’ was produced and the

S.D. Richman et al. / Intra-tumoral heterogeneity of KRAS and BRAF mutation

resulting mutation status was compared with the individual block results. 2.3. Laboratory methods FFPE tumor blocks were retrieved, anonymized and sent to the research laboratory, where staff remained blind to the patients’ identity, treatment and outcomes. Tumor areas were identified on H&E-stained sections then macrodissected from ten 5␮m whole-block sections. Two DNA extraction protocols were used. For FOCUS trial material, DNA was extracted as previously described [14]. For PICCOLO, which requires prospective and hence urgent KRAS testing prior to randomization, the QIAGEN QIAamp DNA Micro kit was used, employing the standard manufacturer’s protocol. The DNA was resuspended in 20␮l water and stored at 4◦ C. DNA was extracted from eight samples using both extraction methods, showing that either methodology could be used to obtain DNA of suitable quality and quantity for analysis. To generate the ‘DNA cocktail’ samples, 5 ␮m sections from multiple individual blocks from the same tumor were macrodissected into a single tube, then the DNA extracted using the QIAGEN DNA Micro kit. Previous validation experiments in the laboratory determined the sensitivity of mutant allele detection to be 5% i.e. pyrosequencing will detect a mutant allele if it comprises as little as 5% of the extracted DNA. The tumor content (%) of the macrodissected region was assessed for each H&E section. Each was given a value within one the categories; 75%. No tumor displayed a tumor content of A mutation, whereas the remaining 5 samples carried a c.34 G > T mutation. It was determined from the pathology report that the patient had 2 synchronous tumors.

Table 3 KRAS and BRAF mutational status of ‘DNA Cocktails’ Sample ID P17 P35 P46 P832 P864 Sample ID F662 P710

Tumour content (%) of block

1st KRAS 12/13 result


2nd KRAS 12/13 result

DNA Cocktail KRAS 12/13 result

T

c.35 G > A c.35 G > A c.35 G > A c.35 G > A c.35 G > A


1st BRAF result


2nd BRAF result


3rd BRAF result

DNA Cocktail BRAF result

25-50